Scitools-iris

==========================

Features added
--------------
* Iris is now an open source project. See `Iris on github <https://github.com/SciTools/iris>`_.
* Hybrid vertical coordinates are implemented with the new :class:`~iris.aux_factory.AuxCoordFactory`.
* More graceful loading of units. See `here <https://github.com/SciTools/iris/pull/17>`_ for more details.

Incompatible changes
--------------------
* The Iris data model is now fully aligned with the `CF data model <https://cf-pcmdi.llnl.gov/documents/cf-conventions/>`_ .
Iris remains file-format independent, as is the underlying CF data model.
* Cube merging has been re-written for the new CF data model with the benefit that some corner cases are now better handled.
Some users may find that their cubes, once merged, now have a smaller total shape and more intelligent handling of dimension coordinate picking.

Deprecations
------------
* None


----------------------------

==========================

Features added
--------------
* Cube trajectory extraction now supports irregular grids (ORCA2) using nearest-neighbour interpolation.
* Support cube arithmetic with masked array data.
* PP loading now utilises improved field-code to standard_name/unit mapping.
* Upgraded to use centrally installed and managed version 2 of PP packing library.

Incompatible changes
--------------------
* The ability to add UK Met Office branding to plots has been moved to
the UK Met Office's internal `monty` library.
* The ability to read colour palettes from the UKMO IDL library has been
moved to the UK Met Office's internal `monty` library.

Deprecations
------------
* None


----------------------------

==========================


Features added
--------------
* Cubes can now have missing data in the form of numpy masked arrays. All plotting and aggregating routines
should now support masked data handling without any user code changes necessary.
* GRIB2 saving now supported for common cases. GRIB loading capabilities further extended.
* NetCDF loading of multidimensional auxiliary data implemented, providing initial support for data with an ORCA grid.
* Extended the number of PP STASH code to standard name mappings.
* PP saving of non ``latitude, longitude`` cubes added.
* Cube merging has been upgraded to support optimisation of ``time``, ``forecast_period`` and
``forecast_reference_time`` dimensionality.
* Added new gallery entry highlighting NetCDF support for rotated pole data.
* NetCDF saving should now be CF-1.5 compliant.
* Custom made colour palettes from the UKMO IDL library can now be loaded as matplotlib colour
palettes (see :func:`iris.clrtools.load_clr`).
* It is now possible to iterate over multiple cubes "in step", ensuring that the metadata is consistent between the
given cubes (see :func:`iris.iterate.izip`).
* Cell methods will now appear in the printout of a cube.
* Supporting software dependency versions have been updated. Of particular note is matplotlib which has gone from version 1.0.1
up to `1.1.0 <https://matplotlib.sourceforge.net/users/whats_new.html#new-in-matplotlib-1-1>`_ . This may have a small impact on
some plot element placements.

Incompatible changes
--------------------
* The call signature for custom loader functions handled via :class:`~iris.io.format_picker.FormatSpecification` has changed.
Previously the signature was ``loader(filename, callback)`` and now looks like ``loader(filenames, callback)`` where ``filenames``
is an iterable of filenames.
* Custom aggregators should now create :class:`iris.analysis.Aggregator` instances rather than defining
a the partial creation of an Aggregate (i.e. ``partial(iris.analysis.Aggregate, ...)`` becomes ``iris.analysis.Aggregator(...)``.
* PP STASH code ingestion is now stricter. In particular handling of partial stash codes (i.e. those without a model or section)
is no longer done, meaning that some cubes which previously had a standard name may no longer have a standard name.
Further work is planned to support field code translation which should provide more robust and accurate PP interpretations.
* For parity with the CF-NetCDF community 'phenomenon_time' has now become 'time'.

Deprecations
------------
* Callback functions to ``iris.load`` and ``iris.load_strict`` should no longer return a value.
Updated documentation can be found at :func:`iris.load`.


----------------------------

==========================

Features added
--------------
* The PP loader now supports PP fields with a header release number of 3.
(As used by the UM version 8.1)

* Initial support for multidimensional coordinates (i.e. those which span multiple
cube data dimensions) with the standard ExplicitCoord class.

* Added cube to CF-netCDF saving.

* Print of a cube now supports multi-dimensional coordinates, and re-categorisation of
cube coordinates into "dimensioned", "scalar" and "other" (i.e. neither dimensioned or
scalar).

* Provide support for CF-netCDF label variables and 1-D auxiliary coordinate variables.

Incompatible changes
--------------------
* The ``cube.coords`` dictionary has been replaced with the ``cube.coord`` and ``cube.coords`` methods
for easier identification of coordinates.
* ``iris.cdm.coords`` has been moved and separated to ``iris.coords and`` ``iris.coord_systems``.
* CF-netCDF global attributes are now put in the cube "attributes" dictionary rather than
as scalar cube coordinates. Additionally any "unused" CF-netCDF variable attributes are also put in the
cube "attributes" dictionary.

Deprecations
------------
* ``cube.axes`` is now deprecated and read-only. A coordinate's dimensions on a cube can be
found with the ``cube.coord_dims`` method. An interim cube convenience property
"axes" is provided to support legacy getter functionality.

* ``cube.axis_length`` is now deprecated. A cube's shape can be determined with ``cube.shape``
and the dimensions to which a coordinate is associated can be determined with ``cube.coord_dims``.


----------------------------

==========================

* Overhaul to NetCDF loading. Iris will now load a far wider variety of CF
compliant NetCDF files.

* Cubes loaded from PP will now have the associated "STASH" code in MSI format
within the cube custom "attributes" dictionary.

* Cubes now have an "attributes" dictionary which can be used for phenomenon
definition - cubes with different attributes are representing fundamentally
different phenomenon.

* Strings can now be used to constrain standard name on load/extract.

* Plotting routines now pick the colour palettes based on the cube to plot.

* Added ability to derive area weights from a shapely geometry.

* Cubes can now be operated on (added, multiplied, etc.) using numpy arrays.

* New example of how to load custom ascii files into Cubes.

* Percentile aggregation added.

* Deferred data loading for Cubes loaded from netCDF.

* Pickle support for cubes. Pickling should be reserved for short term storage and
optimisation purposes only as pickle files will not be compatible across releases.